August 1, 2026
I Wasted 6 Months Using Claude Code Wrong. Here Are the 14 Commands That Changed Everything.
🌟 Special Offer for My Readers 🌟

By Elshad Karimov
5 min read
🌟 Special Offer for My Readers 🌟
Use code:
LEARNWITHMEat checkout on Udemy to get 90% off my courses.
Python Database Course: SQLite, PostgreSQL, MySQL,SQLAlchemy
It was 11:47 PM.
I had spent nearly four hours working on what should have been a simple feature.
My Claude Code session had become a disaster.
Hundreds of messages.
The context window was almost full.
Responses were getting slower.
I was constantly switching between my terminal, browser, GitHub, documentation, and Claude.
Instead of helping me write software…
AI had become another thing I had to manage.
Then another developer looked over my shoulder and asked a simple question.
"Why aren't you using the built-in commands?"
I honestly didn't know they existed.
I thought Claude Code was just a chatbot inside my terminal.
I couldn't have been more wrong.
Over the next few weeks, I completely changed my workflow.
These 14 commands didn't just save me time.
They changed the way I build software.
1. /init — Stop Explaining Your Project Every Time
Every new project started exactly the same.
I'd spend 20 minutes explaining:
- what the application does
- which framework I'm using
- the folder structure
- coding conventions
- project architecture
Only then could I start asking questions.
It felt repetitive.
Then I discovered /init.
Instead of manually giving Claude context, it analyzes the repository and creates a project memory file that documents your codebase.
The next conversation starts with context already available.
No more repeating yourself.
It's one of those commands you run once…
And benefit from every single day afterward.
2. /memory — Teach Claude How You Like to Code
One thing frustrated me.
Claude always wrote perfectly valid code…
Just not the way I preferred.
I'd repeatedly say things like:
- Prefer composition over inheritance.
- Don't use
anyin TypeScript. - Add comments only for complex logic.
- Always include unit tests.
After the tenth time, I realized I was wasting effort.
That's where /memory comes in.
Instead of repeating your preferences in every conversation, store them once.
From then on, Claude adapts to your coding style instead of forcing you to adapt to its defaults.
It feels less like using AI…
And more like working with a teammate who already knows your habits.
3. /pr_comments — Never Miss Review Feedback Again
Pull requests can become overwhelming.
One reviewer comments on naming.
Another asks for better error handling.
Someone else suggests changing the architecture entirely.
Normally, I'd jump between GitHub and Claude trying to keep everything in sync.
Then I found /pr_comments.
It loads the review comments directly into your session.
Now Claude understands exactly what reviewers requested before we start making changes.
No forgotten comments.
No unnecessary tab switching.
Just a smoother review process.
4. /btw — Stop Breaking Your Flow
You're halfway through building an authentication system.
Then suddenly a question pops into your head.
"What's the difference between JWT expiration and refresh tokens?"
Normally you'd Google it.
Or ask Claude.
Either way…
Your main task gets interrupted.
/btw solves this beautifully.
It lets you ask side questions without derailing your current conversation.
It sounds like a tiny feature.
In reality…
It's one of the biggest productivity improvements I've experienced.
5. /compact — My Favorite Command
Every long session reaches the same point.
Responses slow down.
The context window fills up.
You wonder whether it's time to start over.
Before, I'd open a brand-new session.
Lose all my context.
Spend another 20 minutes explaining everything again.
Now I type one command.
/compact
Claude summarizes the important parts of the conversation, removes unnecessary history, and keeps the project moving.
The first time I used it, I genuinely wondered why nobody had told me about it sooner.
6. ! — Run Terminal Commands Without Leaving Claude
This one completely changed my workflow.
Instead of switching windows every minute…
I simply type:
! git status! git statusor
! npm test! npm testor
! npm install! npm installThe output appears directly inside the conversation.
No copy-paste.
No context switching.
No breaking concentration.
It's a small improvement that saves dozens of tiny interruptions every day.
7. /cost — Know Your Usage Before It Becomes a Problem
I never thought about token usage.
Until I hit the limit.
Suddenly Claude started forgetting earlier parts of the conversation.
I had no idea why.
/cost shows your current usage, remaining context, and other session statistics.
Think of it as the fuel gauge for your AI session.
Instead of running out unexpectedly…
You know exactly when it's time to optimize.
8. /model — Pick the Right Brain for the Job
Not every task needs the same model.
Sometimes I'm solving a difficult architectural problem.
Sometimes I just want a quick regex.
With /model, switching between models takes seconds.
Heavy reasoning?
Choose a more capable model.
Fast iteration?
Switch to a lighter one.
Learning when to change models made Claude feel far more flexible than I originally thought.
9. /fast — Speed Matters
There are moments when perfect answers aren't necessary.
You're brainstorming.
Trying ideas.
Refactoring.
Experimenting.
In those moments…
Speed beats perfection.
/fast gives quicker responses, making rapid iteration much smoother.
Then, once the implementation is complete, I switch back for deeper reviews.
10. /review — Like Having a Senior Engineer Beside You
Before every merge, I now run one command.
/review
It doesn't just check formatting.
It looks for:
- logic bugs
- performance issues
- security risks
- missing edge cases
- code quality improvement
Sometimes it catches things I completely missed.
Other times it simply confirms I'm on the right track.
Either outcome is valuable.
11. /clear — A Fresh Start Without Losing Everything
Sometimes conversations simply go off the rails.
Wrong assumptions.
Wrong direction.
Too much noise.
Instead of closing Claude completely…
/clear resets the conversation while keeping your overall setup intact.
It's like cleaning your desk without moving offices.
12. /doctor — The First Command to Run When Something Feels Broken
If Claude behaves strangely…
Don't panic.
Run /doctor.
It checks your setup, configuration, permissions, and environment before you waste time debugging something that isn't your code.
It's surprisingly useful.
13. /terminal-setup — Connect Claude to Your Workflow
One of the biggest frustrations used to be copying terminal output back into Claude.
Again.
And again.
And again.
/terminal-setup bridges that gap.
Once configured, Claude works much more naturally alongside your terminal instead of feeling like a completely separate tool.
14. /help — The Command Everyone Forgets
Ironically…
The most useful command might be the simplest one.
Whenever I forget what Claude can do…
I type:
/help
It reminds me of features I haven't used in weeks.
Even after months of using Claude Code, I still discover commands I'd forgotten existed.
Final Thoughts
For six months, I thought Claude Code's biggest feature was generating code.
I was wrong.
Its biggest feature is improving how you work.
The difference isn't just better prompts.
It's understanding that Claude Code is an entire development environment, not just an AI chatbot.
If you're only asking it to write functions…
You're probably using a fraction of its capabilities.
Learn these 14 commands.
Use them for a week.
Then try going back to your old workflow.
I couldn't.
And I don't think you will either.
🚀
If you found them helpful:
- Show some love with a 50-clap
- Drop a comment.
- Share your favorite part!
Discover more in my online courses at Appmillers
Connect on LinkedIn: Elshad Karimov
Follow on X (Twitter) : Elshad Karimov